1. Field of the Invention
The present invention relates to a semiconductor device, a method of manufacturing the semiconductor device, and a display apparatus incorporating the semiconductor device, and more particularly to a semiconductor device having a low-resistance bus interconnect, a method of manufacturing the semiconductor device, and a display apparatus incorporating the semiconductor device.
2. Description of the Related Art
Liquid crystal displays (LCDs) are incorporated in various portable devices including cellular phone sets, PDAs (Personal Digital Assistants), notebook personal computers, etc. because the LCDs are low in profile and light in weight. For medium- and large-size LCDs such as television LCDs to find wider use, however, they need to be lower in cost, particularly, the packaging cost for driver semiconductor devices for driving liquid crystal cells needs to be lower.
Heretofore, driver semiconductor devices for medium- and large-size LCDs have been packaged according to the TAB (Tape Automated Bonding) process. The TAB process is a process of connecting one side of output terminals of a TCP (Tape Carrier Package), which comprises a single semiconductor device mounted on a single flexible board, to a signal input section of an LCI panel board. Medium- and large-size LCDs are required to be combined with ten to twenty driver semiconductor devices. The individual driver semiconductor devices have to be successively supplied with input signals and power supply voltages for driving liquid crystal cells from outside of the LCD. Therefore, a printed circuit board having a plurality of bus interconnects thereon is connected to one side of input terminals of each TCP.
The input signals and power supply voltages are supplied from the bus interconnects on the printed circuit board through bus interconnects on the TCPs to the individual drive semiconductor devices. To the printed circuit board, there is connected a flexible board, other than the TCPs, for supplying the input signals and power supply voltages.
The medium- and large-size LCDs need the ten to twenty driver semiconductor devices for the following reasons: An LCD comprises a two-dimensional array of pixels arranged in rows and columns which extend perpendicularly to each other. At present, the matrix of pixels that makes up medium- and large-size LCDs is primarily an XGA (Extended Graphics Array) having vertically spaced 768 rows and horizontally spaced 1024 columns.
For displaying color images, an XGA panel is made up of vertically spaced rows of 768 dots and horizontally spaced columns of 3072 dots because three dots corresponding to three primaries are arranged in a horizontal array as one pixel. On active-matrix liquid crystal panels, each dot is associated with a thin-film transistor (TFT). A voltage is applied to a gate line extending along a row to turn on the TFTs that are connected to the row, and a signal voltage is applied from a source line extending along a column to one of the TFTs that are turned on, energizing the corresponding dot to display an image.
For displaying an image on an XGA panel, therefore, it is necessary for one or more gate driver semiconductor devices to have output terminals corresponding to 768 gate lines and for one or more source driver semiconductor devices to have output terminals corresponding to 3072 source lines. If a gate driver semiconductor device having 768 output terminals could be constructed as a single semiconductor chip and a source driver semiconductor device having 3072 output terminals could be constructed as a single semiconductor chip, then an XGA panel requires only two semiconductor devices (one gate driver semiconductor device and one source driver semiconductor device). However, such semiconductor devices are not presently available in the art because their chip size is too large.
According to the present practical application of packaging technology, the minimum pitch that allows packaging at production sites is of about 40 μm. That is, for achieving reliable connections through outer lead bonding and inner lead bonding, each terminal needs a width of 40 μm or more. If a source driver semiconductor device having 3072 output terminals is constructed as a single chip, then the dimension of a longitudinal side of the chip is 40 μm×3072=123 mm, which makes the chip too large to be practically formed on a single-crystal silicon substrate. For this reason, general driver semiconductor devices available in the market have 200 to 400 output terminals. At present, most XGA panels have a total of 11 driver semiconductor devices including 3 gate driver semiconductor devices each having 256 output terminals and 8 source driver semiconductor devices each having 384 output terminals.
Since ten to twenty TCPs, one or two flexible boards, and one or two printed circuit boards are required according to the TAB process, the cost of the parts and the cost required to package those parts are responsible for an increase in the manufacturing cost of LCDs.
One way of reducing the packaging cost of LCD driver semiconductor devices is a COG (Chip On Glass) process that has begun to be used in practical applications instead of the TAB process (see, for example, Japanese laid-open patent publication No. 2003-100982). The COG process is a process of mounting a driver semiconductor device as a bare chip directly on an LCD panel board without the need for an TCP. Specifically, the circuit surface of a driver semiconductor device is pressed with heat against the LCD panel board. At this time, an adhesive tape of synthetic resin with electrically conductive particles dispersed therein is sandwiched between protrusive output terminals, ten to twenty μm high, on the circuit surface of the driver semiconductor device and input terminals on the LCD panel board. Only those areas of the adhesive tape which are pressed between the terminals are rendered electrically conductive, thus electrically interconnecting the terminals. Even when the COG process is employed, bus interconnects are required to successively connect individual driver semiconductor devices. To meet such a requirement, Japanese laid-open patent publication No. 2003-100982 discloses a process of forming parallel bus interconnects on an LCD panel board and a driver semiconductor device to lower the interconnect resistance for minimizing flexible boards and printed circuit boards, thereby to allow an LCD to be assembled at a low cost.
Simply replacing the TAB process with the COG process is unable to reduce the number of driver semiconductor devices installed on an LCD panel board through the COG process is expected to reduce the number of flexible boards and printed circuit boards. According to a proposed process of the number of installed driver semiconductor devices, thin-film transistors are formed on two slender glass boards which are as long as two respective perpendicular sides of an LCD panel board, producing a gate driver semiconductor device and a source driver semiconductor device, and those semiconductor devices are mounted on the LCD panel board in the same manner as the COG process, thereby assembling a liquid crystal display unit at a lower cost (see, for example, Japanese patent No. 3033123). As described above, if each of the gate and source driver semiconductor devices is fabricated on a single-crystal silicon substrate, it is unpractical to form a driver semiconductor device having a few thousand output terminals on a single chip due to dimensional limitations on silicon substrates. However, glass substrates are much less subject to such dimensional limitations because they can be produced in sizes sufficiently larger than silicon substrates. If a driver semiconductor device is formed on a glass substrate, then it is possible to fabricate a driver semiconductor device of greater dimensions which has more output terminals, and it is expected that LCDs can be assembled at a lower cost using such a driver semiconductor device.
The driver semiconductor device on the glass substrate is still problematic, however, in that on the driver semiconductor device of greater dimensions which has more output terminals, voltage drops developed in bus interconnects manifest themselves, tending to increase variations in voltages supplied to a plurality of drive circuit units for supplying voltages to gate lines and source lines on an LCD panel board. For example, if an image is to be displayed on an LCD with a single large-size driver semiconductor device, then voltage drops developed in power supply bus interconnects in the driver semiconductor device are liable to produce a large difference between the voltage supplied to the drive circuit unit that is closest to a power supply feed point and the voltage supplied to the drive circuit unit that is farthest from the power supply feed point. Such a large difference between the voltages supplied to individual drive circuit units fails to keep a desired level of displayed image quality because of a gradation shift displayed on the LCD, and, in some cases, prevents pixels connected to the driver circuit from being displayed.
Therefore, if a large-size driver semiconductor device which is capable of driving medium- and large-size LCDs alone is to be used, then either low-resistance bus interconnects have to be formed in the driver semiconductor device or on the LCD panel board, or an LCD panel board with low-resistance interconnects have to be additionally connected to the driver semiconductor device.
For example, a bus interconnect having a resistance of 5 Ω/m per unit length is required to realize a driver semiconductor device which has 3000 drive circuit units formed across a width of 250 mm. It is assumed that a power supply feed point is disposed at one end of the driver semiconductor device, and the bus interconnect extends from that one end to the other of the driver semiconductor device. The resistance R per unit length of the bus interconnect can be expressed as R=ρ/A, where p represents the volume resistivity of the bus interconnect (Ω·m) and A represents the cross-sectional area of the bus interconnect (m2). If the bus interconnect is made of Au, then since the volume resistivity of Au at room temperature is about 2.3×10−8 Ω·m, the bus interconnect is required to have a cross-sectional area of 4600 μm2 in order for the bus interconnect to have the resistance R of 5 Ω/m. Therefore, if a bus interconnect having a width of 200 Ωm is formed and its resistance R is to be equal to or less than 5 Ω/m, then the bus interconnect has to have a thickness of 23 μm or greater.
Generally, interconnects in driver semiconductor devices or on LCD panel boards are formed by vacuum evaporation or sputtering, and are difficult to have low resistances. The thickness of a film that can be deposited by vacuum evaporation or sputtering is limited to about 1 μm at maximum. Consequently, if an aluminum interconnect having a width of 500 μm and a thickness of 1 μm, for example, is formed, on the assumption that the volume resistivity of aluminum is 2.78×10−8 Ω·m, then the aluminum interconnect has a high resistance of about 56 Ω/m per unit length. When a thicker aluminum film is formed, the substrate or board may be warped, the film may be peeled off, and the surface configuration may be impaired, due to internal stresses of the film, making it difficult to perform subsequent manufacturing processes.
The thicker bus interconnects present an obstacle to efforts to package a semiconductor device in the same manner as the COG process. This is because packaging a semiconductor device in the same manner as the COG process requires that structural bodies on the semiconductor device substrate be essentially of the same height.